Professional Documents
Culture Documents
Characteristics and Prediction of Sound Level in Extra-Large Spaces
Characteristics and Prediction of Sound Level in Extra-Large Spaces
Applied Acoustics
journal homepage: www.elsevier.com/locate/apacoust
A R T I C L E I N F O A B S T R A C T
Keywords: This paper aims to examine sound fields in extra-large spaces, which are defined in this paper as spaces used by
Extra-large space people, with a volume approximately larger than 125,000 m3 and absorption coefficient less than 0.7. In such
Sound field spaces inhomogeneous reverberant energy caused by uneven early reflections with increasing volume has a
Sound pressure level significant effect on sound fields. Measurements were conducted in four spaces to examine the attenuation of the
Prediction
total and reverberant energy with increasing source-receiver distance, which was then validated by the simu-
lations with image-source method. Results show that the reasons for the total energy’s exponential decrease are
not only the direct sound, but also the reverberant energy. The prediction difference of total sound pressure level
(SPL) between classical formula and the image-source method increases with the volume and decreases with the
surface absorption, based on which a critical line separating extra-large spaces from large ones is proposed.
Moreover, a newly modified model based on the importance of first reflection from floor is proposed, showing
more advantages of sound level prediction in extra-large spaces.
⁎
Corresponding author at: School of Architecture, University of Sheffield, Sheffield S10 2TN UK.
E-mail addresses: j.kang@sheffield.ac.uk, j.kang@hit.edu.cn (J. Kang).
https://doi.org/10.1016/j.apacoust.2017.12.023
Received 24 June 2017; Received in revised form 10 December 2017; Accepted 20 December 2017
0003-682X/ © 2017 Elsevier Ltd. All rights reserved.
C. Wang et al. Applied Acoustics 134 (2018) 1–7
point of the floor at a height of 1.5 m . The receiver height was 1.2 m .
Each point was measured three times to obtain the average value. ISO-
3382 standard was followed in the equipment selection and measure-
ment process. For the sake of brevity, only the average of three octave
bands (500 Hz,1 kHz and 2 kHz) is presented in this paper. The SPL at all
points were more than 15 dB above the background noise in the four
cases.
2.1. On-site measurements Basic characteristics of sound energy distribution were explored in
the measurements of four cases, as shown in Fig. 3. A near-exponential
On-site measurements were carried out in four cases, with volumes trend both appears in the distribution of total and reverberant energy in
varied from 7000 to 190,000 m3, representing ordinary to large space all four cases and this phenomenon is more obvious near the source. In
respectively, as shown in Fig. 1 and Table 1. The shape and acoustic the area far from the source, the attenuation gradually becomes linear.
performance of surfaces in four cases are different but all of them are For each case, a polynomial regression is performed of the total sound
centralized. Other information is shown in Fig. 2, and the RT in this level data as a function of source-receiver distance. The R2 of regression
Table 1 is the average reverberation time of three octave bands in four cases varies from 0.87 to 0.92, which indicates that spatial
(500 Hz,1 kHz and 2 kHz). difference between the points at the same source-receiver distance ex-
The SPL values were measured using an omni-directional source ists in a small range.
with a pink noise signal and calibrated sound level meters. The source In order to further determine the spatial distribution of the re-
points were arranged at the usual performance point or near the central verberation energy in the space of larger volume and greater absorption
Table 1
Geometry and information of four cases.
Case Name Volume (m3 ) Seats RT (s) Receiver points Background noise (dBA) Temperature (°C ) Humidity (%)
2
C. Wang et al. Applied Acoustics 134 (2018) 1–7
coefficient, a series of cubes were simulated by an image-source relationship between the dash straight line and other curves in Fig. 4,
method, the results of which are presented in Fig. 4. In the cubic space the direct energy in the space with larger surface absorption occupies a
with a volume of 1000 m3 , the reverberant energy attenuates slightly greater proportion in total energy and shows more exponential trends.
with the increase of source-receiver distance. The difference range of Moreover, the curves in Fig. 4(a) are steeper than those in Fig. 4(b)
reverberant energy in the space is 1.1 dB and 2.5 dB while the absorp- which means that the reverberant energy also appears to be a more
tion coefficient is 0.3 and 0.7, respectively. This phenomenon is exponential trend with the growth of surface absorption. In each figure,
common in ordinary large spaces, such as auditoriums and small indoor this exponential trend is also more evident with the growth of volume.
stadiums. However, with the increase in volume and absorption coef- In this way, there is a larger distance range with exponential behaviour
ficient, reverberant SPL shows a significant exponential decreasing for larger room volumes and greater absorption coefficients.
trend. From Fig. 4, it is obvious that the curves become closer to the
straight direct-energy line, since logarithmic abscissa axis is used. 4. Prediction model
For large cubic spaces with side lengths of 100 m or more, the SPL
profiles in Fig. 4 can be divided into three sections. The first part is the The applicability of the classical model is first studied in this sec-
linear attenuate curve near the source and in different spaces those tion. Then a new prediction model is proposed theoretically and vali-
curves are always within about 6 m from the source. Then the second dated with the measurements.
part shows an exponential trend in the remaining near-source area.
More exponential trends appear with the increase of volume and the
4.1. Applicability of classical model
reverberant SPL curves of different spaces in the two figures move down
and gradually approaches the first reflection from floor (FRFF) curve.
The classical model [19] assumes that the reverberation sound en-
The last part backs to the linear attenuate trend again in the area far
ergy is evenly distributed in a space, hence, the total SPL is
from the source. This linear pattern is due to the domination of other
reflections, which still have a slight attenuation with distance because W 4W
L = 10log10 ⎛ + ⎞ + 120
the energy of FRFF decreases with the source-receiver distance. This ⎝ 4πr 2 R ⎠ (1)
three-part phenomenon in a large space is different from the hypothesis
of classical model or other models [2,22]. where W is the sound source power, r is the source-receiver distance,
In terms of the effect of surface absorption, according to the relative and R is the room constant.
As the volume becomes larger and the sound absorption coefficient
3
C. Wang et al. Applied Acoustics 134 (2018) 1–7
W 4W −13.82r
LB = 10log10 ⎛ + ·e cTE ⎞ + 120
⎝ 4πr 2 R ⎠ (2)
0 tD tf tc
Fig. 5. The prediction difference of the total energy (dB) obtained by the classical model TIME
and the image-source method (classical minus image-source method). Only the max ab-
solute value of the difference was recorded in one space. Fig. 6. Different stages of the energy decay.
4
C. Wang et al. Applied Acoustics 134 (2018) 1–7
1.0 where ΔH is the height difference between the source and receiver
points.
0.5
r 2 = (ΔH )2 + Lh2 (4)
0
2 4 6 8 10 12 14 16 18 where Lh is the horizontal distance between the source and receiver
Source-receiver Distance (m) points.
Fig. 7. Path difference between the direct sound and the first reflected sound from the
L2 = (2Hs−ΔH )2 + Lh2 (5)
floor.
where L is the path length of FRFF.
Source Receiver According to Eqs. (3)–(5), the path length of FRFF can be obtained
by
r
ΔH
L= 4Hs Hr + r 2 (6)
Hs
Hr
As the FRFF is considered, the total energy can be divided into three
parts, which are the energy of direct sound, the FRFF and other sub-
L sequent reverberant sounds. So the total SPL can be obtained by
Hs
W W (1−αf ) 4W −13.82L ⎞
L′ = 10log10 ⎛⎜ + + ·e cT ⎟ + 120
Lh ⎝ 4πr 2 4πL2 R ⎠ (7)
Image source where αf is the absorption coefficient of the floor; T is the reverberation
time measured or calculated. Adding Eqs. (6) and (7), L′ will be gen-
Fig. 8. The geometry of sound source and receiver points based on the image method. erated as:
5
C. Wang et al. Applied Acoustics 134 (2018) 1–7
Table 2
Differences of measured and predicted values by three models in four cases.
Dangxiao auditorium −1.7 3.49 1.46 −2.09 0.58 0.72 −0.94 1.55 0.66
HIT indoor stadium −2.02 2.52 1.36 −2.36 0.98 0.9 −1.2 1.32 0.73
TJU indoor stadium −3.38 2.72 1.95 −3.5 0.5 1 −1.38 1.33 0.66
Guian Circus Show −2.4 0.52 1.03 −2.71 −0.34 0.77 −0.7 1.49 0.71
Fig. 10. Difference in the results of models and computer simulation (models minus computer simulation). The dotted lines represent the Barron model, and the solid lines represent the
classical model. The dashed and grey solid lines represented the modified model using Sabine and Eyring reverberation time respectively. (a), (b), and (c) were in the space with volumes
1000 m3,125,000 m3 and 125,000,000 m3 respectively when the surface coefficient was 0.1. (d), (e), and (f) were in the space with volumes 1000 m3,125,000 m3 and 125,000,000 m3 re-
spectively when the surface coefficient was 0.5.
4.3. Validation of the new model source-receiver distance, as shown in Fig. 10. When the volume is
1000 m3 , all the three models have little differences, which are not
The modified model is applied to the four measured cases to obtain larger than 1 dB. When the volume is 125,000 m3 and 125,000,000 m3 ,
further validation, and the classical and Barron models are also used for both curves of the classical and Barron models show a deep trough at
comparison, as shown in Fig. 9. In the four cases, the absorption coef- about 4–20 m , which means a different degree of underestimation in the
ficient is calculated by the Eyring formula based on the reverberation near-source area. This trough is mainly due to the FRFF and its re-
time measured, and the sound power level of the source is deduced lationship with other energy as in Fig. 4, it is proportional to the volume
from the SPL measured at 1 m from the source, since the direct energy and inversely proportional to the absorption coefficient. When the
here is about 9.5 dB larger than the first reflect energy from the floor surface coefficient is 0.1 and the volume is 125,000,000 m3 , the max
using the spherical formula. difference between two existing models and computer simulation is
As can be seen in Fig. 9, the results show that the classical model 2.7 dB.
underestimated the energy in the near field, within about 10–20 m , and In Fig. 10, as the volume increases, the prediction difference of all
overestimated it in the far field. By adding a distance-dependent at- the four models becomes greater generally and the difference curves
tenuation factor to reverberation energy, Barrons model yields very show more spatial fluctuation with increasing source-receiver distance.
accurate predictions in the far field but still underestimated the energy This may suggest more theoretical limitations of statistical models in
in the near field. However, as the volume becomes larger and the first the extra-large spaces as the sound rays become more difficult to reach
reflection energy from the floor has more advantages than other re- the statistical average with the increasing volume. Different from the
flections, the accuracy is better using the modified model. The differ- two existing models, the prediction difference of the two modified
ences between the measurements and the modified model are sig- models decreases with the increasing source-receiver distance. This is
nificantly reduced. The maximum difference in the four cases is mainly due to the fact that the starting point of the calculation for the
between 1.38 and 1.55 dB and the standard deviation is 0.66–0.71 dB , as reverberant energy in the modified models is the latest among those
shown in Table 2. models which results in less reverberant energy at the far-source point.
Further information is obtained from the simulation. The difference With regard to the use of Sabine or Eyring reverberation time, it is
among the three models (classical, Barron, and the modified) and difficult to have the exact evidence to determine by computer simula-
computer simulation shows different trends with the increase of the tion or measurements because most of the difference between the two
6
C. Wang et al. Applied Acoustics 134 (2018) 1–7
0.9 not be used in the situation where the sound source, or the floor on
-0.1 which the first-order sound reflects, cannot be seen at the receiver
0.8 -0.7
Coefficient of absorption
point. The applicability in the situation where the source or receiver is
0.7 -0.9 -0.3 far from the floor is also uncertain.
-1
0.6 Acknowledgements
-1.1 -0.5
0.5
We acknowledge Python Software Foundation and Continuum
-0.7
0.4 Analytics for their open-source software python and anaconda.
Funding: This work was supported by the National Natural Science
0.3 -0.9 -0.9 Foundation of China [grant numbers 51378139 and 51178300].
0.2 -0.7 -1.1 References
0.1
100 200 300 400 500 [1] Fujikawa T, Aoki S. An escape guiding system utilizing the precedence effect for
3
V (m) evacuation signal. Proceedings of meetings on acoustics, vol. 19. Acoustical Society
of America; 2013. p. 030055.
Fig. 11. The prediction difference of the modified model and the image-source method [2] Barron M. Theory and measurement of early, late and total sound levels in rooms. J
(modified minus image-source method). Acoust Soc Am 2015;137(6):3087–98.
[3] Aretz M, Orlowski R. Sound strength and reverberation time in small concert halls.
Appl Acoust 2009;70(8):1099–110.
formulas are within 1 dB as shown in Fig. 10, which still requires further [4] Jeon JY, Kim YH. Design of sound diffusion in concert halls using scale models. J
Acoust Soc Am 2010;127(3):1752.
study. [5] Soeta Y, Ito K, Shimokura R, Sato S-I, Ohsawa T, Ando Y. Effects of sound source
More details of the difference between the modified model and the location and direction on acoustic parameters in Japanese churches. J Acoust Soc
image-source method are shown in Fig. 11. The modified model shows Am 2012;131(2):1206–20.
[6] Zamarreño T, Girón S, Galindo M. Acoustic energy relations in mudejar-gothic
a better applicability in prediction and the accuracy of this new model churches. J Acoust Soc Am 2007;121(1):234–50.
is insensitive to volume, which shows more applicability in extra-large [7] Lewers T, Anderson J. Some acoustical properties of St Paul’s Cathedral, London. J
spaces. The max difference of total SPL appears when the surface ab- Sound Vib 1984;92(2):285–97.
[8] Joyce W. Exact effect of surface roughness on the reverberation time of a uniformly
sorption is 0.5. absorbing spherical enclosure. J Acoust Soc Am 1978;64(5):1429–36.
[9] Miles R. Sound field in a rectangular enclosure with diffusely reflecting boundaries.
5. Conclusion J Sound Vib 1984;92(2):203–26.
[10] Hodgson M. When is diffuse-field theory accurate? Can Acoust 1994;22(3):41–2.
[11] Galaitsis AG, Patterson WN. Prediction of noise distribution in various enclosures
For SPL-prediction purposes, if 1 dB is accepted as the criterion, the from free-field measurements. J Acoust Soc Am 1976;60(4):848–56.
space with the absorption coefficient is less than 0.7 and the volume is [12] Kuttruff H. Stationary propagation of sound energy in flat enclosures with partially
diffuse surface reflection. Acta Acust United Acust 2000;86(6):1028–33.
large than 125,000 m3 is suggested to be called extra-large space based
[13] Kang J. Acoustics in long enclosures with multiple sources. J Acoust Soc Am
on the results of computer simulation in the current paper. 1996;99(2):985–9.
When a space volume becomes larger, the diffuse field assumption is [14] Picaut J, Simon L, Polack J-D. Sound field in long rooms with diffusely reflecting
more and more inapplicable not only for the total energy but also for boundaries. Appl Acoust 1999;56(4):217–40.
[15] Tang S, Piippo K. Sound fields inside street canyons with inclined flanking building
the reverberant energy. The three-stage phenomenon was found in the façades. Proceedings of meetings on acoustics 161ASA, vol. 12. ASA; 2011. p.
attenuation of reverberant energy by using the image-source method 040004.
and was validated in the following investigation. The first part is the [16] Anderson J, Bratos-Anderson M. Acoustic coupling effects in St Paul’s Cathedral,
London. J Sound Vib 2000;236(2):209–25.
linear attenuate curve near the source and then the second part shows [17] Martellotta F. Identifying acoustical coupling by measurements and prediction-
an exponential trend in the remaining near-source area. The last part models for St. Peters Basilica in Rome. J Acoust Soc Am 2009;126(3):1175–86.
reverts to the linear attenuate trend again in the area far from the [18] Sendra JJ. Computational acoustics in architecture. Wit Press; 1999.
[19] Kuttruff H. Room acoustics. CRC Press; 2009.
source. The importance of the first reflection from the floor in an extra- [20] Gibbs BM, Jones D. A simple image method for calculating the distribution of sound
space is found to be the main reason of the three-stage attenuation. pressure levels within an enclosure. Acta Acust United Acust 1972;26(1):24–32.
A new model to predict the SPL attenuation in extra-large spaces is [21] Lehmann EA, Johansson AM. Prediction of energy decay in room impulse responses
simulated with an image-source model. J Acoust Soc Am 2008;124(1):269–77.
proposed based on the importance of the first reflection from the floor. [22] Cirillo E, Martellotta F. Sound propagation and energy relations in churches. J
In this model the energy of the first reflection from the floor can be Acoust Soc Am 2005;118(1):232–48.
calculated separately rather than being included in the approximate [23] Vorländer M. Revised relation between the sound power and the average sound
pressure level in rooms and consequences for acoustic measurements. Acta Acust
formula of the reverberant energy, and its accuracy and applicability
United Acust 1995;81(4):332–43.
were validated by two methods. In computer simulation, the advantage [24] Berardi U, Cirillo E, Martellotta F. A comparative analysis of acoustic energy models
of the new model in the near-source area is indicated, and in four for churches. J Acoust Soc Am 2009;126(4):1838–49.
measured cases, the max difference is from 1.38 dB to 1.55 dB and the [25] Polack J-D. Playing billiards in the concert hall: the mathematical foundations of
geometrical room acoustics. Appl Acoust 1993;38(2):235–44.
standard deviation of which is from 0.66 dB to 0.71 dB . This model could